0
0
Cutting-edge artificial intelligence (AI) is rapidly advancing to the point where it rivals the clinical knowledge and reasoning skills of specialist eye doctors, according to a study led by the University of Cambridge.
The study, published in PLOS Digital Health, evaluated the performance of GPT-4—a state-of-the-art large language model—against doctors at various career stages, including unspecialized junior doctors, and trainee and expert eye doctors. Through a series of 87 patient scenarios involving specific eye problems, participants were tasked with diagnosing or advising on treatment options.
Results revealed that GPT-4 outperformed unspecialized junior doctors, akin to general practitioners in eye knowledge, and achieved comparable scores to trainee and expert eye doctors. While the top-performing doctors scored slightly higher, the findings underscored the potential of AI to augment clinical decision-making in healthcare.
Lead author Dr. Arun Thirunavukarasu, formerly of the University of Cambridge and now an Academic Foundation Doctor at Oxford University Hospitals NHS Foundation Trust, highlighted the practical applications of AI in triaging patients with eye issues.
“With further development, large language models could provide valuable assistance in triaging patients, determining which cases require immediate specialist attention, and offering management suggestions,” said Dr. Thirunavukarasu.
The study emphasized the importance of large volumes of clinical text to refine and develop AI models. Unlike previous studies comparing AI to examination results, this research directly assessed AI against the on-the-spot knowledge and abilities of practicing doctors, providing a more realistic evaluation.
Furthermore, the research tested various AI models, including GPT-3.5, PaLM2, and LLaMA, with GPT-4 emerging as the most accurate responder. GPT-4, known for powering the online chatbot ChatGPT, has garnered attention for its performance in medical school examinations and its ability to provide accurate and empathetic responses to patient queries.
“While AI continues to advance rapidly, the role of doctors in patient care remains paramount,” noted Dr. Thirunavukarasu. “Ultimately, patients should have the autonomy to decide whether they prefer AI involvement in their healthcare journey.”
As AI continues to evolve, the study highlights the need for ongoing research to assess its capabilities and limitations. With advancements in AI technology, including newer models released since the study, the potential for AI to augment clinical decision-making in various medical specialties, including ophthalmology, continues to expand.
The findings underscore the transformative potential of AI in healthcare and the importance of collaborative efforts between AI developers, healthcare professionals, and patients to harness its benefits effectively.
